Liver Electronic Offering Platform with Artificial intelligence-based Devices

Project Information

LEOPARD

Grant agreement ID: 101080964

DOI

10.3030/101080964

EC signature date 14 October 2023

Start date 1 November 2023

End date 31 October 2028

Funded under

• Health

Total cost

€ 6 530 108,55

EU contribution

€ 6 320 857,30

Coordinated by

ASSISTANCE PUBLIQUE HOPITAUX DE PARIS
France

Periodic Reporting for period 2 - LEOPARD (Liver Electronic Offering Platform with Artificial intelligence-based Devices)

Reporting period: 2025-05-01 to 2026-04-30

Summary of the context and overall objectives of the project

Liver transplantation (LT) is a life-saving procedure for decompensated cirrhosis (DC) and hepato-cellular carcinoma (HCC). Its efficacy is hampered by the risk of death/drop-out on the Wait List (WL). This risk is driven by organ shortage and is mitigated by organ offering schemes. According to a sickest first policy, offering schemes prioritize LT candidates with the highest risk of dying, as assessed by predictive models. To drive allocation, Organ Sharing Organizations (OSOs) use a 20-year-old model, the MELD, predicting mortality in DC but not in HCC. Because of a dramatic increase in % of HCC candidates (40% vs 10% in early 20ties), MELD schemes are increasingly inaccurate, with persisting 15 to 30% mortality in countries with low/medium donation rate. This scenario, together with advances in prognosis in DC and HCC candidates and statistics, prompts LT community to look for new algorithms to refine offering schemes. To address this issue, key European LT stakeholders including OSOs, LT experts, Statisticians, Research Labs and SME joined LEOPARD. LEOPARD propose to design and validate an AI-based LEOPARD predictive algorithm outperforming current allocation MELD-based models by better stratifying patients on the risk of mortality, to be proposed OSOs to drive allocation as well as LEOPARD calculators available for professional for assistance in complex decision-making processes. We expect that these computational tools improve candidates’ outcomes, with more patients transplanted on time and significant reduction in the risk of death on the WL. Also, adoption of LEOPARD tools should result in harmonization of European heterogeneous prioritization schemes, and in a significant reduction in disparities of access to LT, a major objective pointed out by the European Commission. LEOPARD should place Europe in leading position for organ offering schemes.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

LEOPARD project first identified from the literature a set of variables with potential predictive value to serve as the foundation for training advanced 2nd-generation AI-based predictive models. These include both routinely collected data from national OSO systems and a set of additional, clinically relevant variables selected by expert clinicians. Two tailored datasets were created: one for patients with DC and other end-stage liver diseases, and another for patients with HCC in order to obtain clinical predictors relevant to each patient population.

On this ground, the LEOPARD Training and Validation Cohort Study (TVDCS) was officially launched in February 2025.A major milestone (MS 2) was achieved in the Training and Validation Data Collection Study (TVDCS – NCT06675604) by June 2026, with 54 clinical sites open to recruitment and 1,000 patients enrolled across the seven participating European countries. In parallel, substantial efforts have been dedicated to ensuring the completeness, and overall quality of the collected data, enabling the start of the first interim statistical analysis. Preliminary results identified at baseline several innovating predictors of mortality/drop out with high predictive value offering the potential of a robust foundation for the development of the second-generation LEOPARD predictive models with improved patients’ stratification.

Significant progress was also made in defining and implementing the processes required to incorporate new predictive variables into the OSOs' software platforms (Milestone 4).

Regarding the second clinical LEOPARD study, PVC1 (NCT06723275), regulatory approval was obtained in France on September 2025, and three clinical sites are currently open for patient recruitment. For the remaining French centers, site initiation is currently underway. Outside France, the first regulatory approval was obtained for the national coordinating center in Italy in mid-June. In parallel, regulatory submissions and site agreement negotiations have progressed substantially in the remaining participating countries, with all participating centers expected to be open for recruitment by January 2027. In addition, the final version of the eCRF has been validated and is fully operational. First enrolments in PVC1 are expected early July 2026.

Another major milestone was the initiation of HCC imaging data transfer to establish the LEOPARD Imaging Repository for radiomic analyses. The first imaging datasets have been successfully uploaded within the framework of the LEOPARD TVDCS study, and the repository is now fully operational and ready to receive imaging data from the LEOPARD PVC1 study. The implementation of this repository required an extensive regulatory approval process and full GDPR compliance, a process currently achieved.

Overall, collaboration among clinicians, AI researchers, data scientists, and patient organizations continues to be a key driver of the LEOPARD project. The project has successfully transitioned from the setup phase to the technical implementation and development phase, marking an important step towards the development and validation of second and next-generation predictive algorithms in liver transplant candidates.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

LEOPARD is the first European initiative to develop AI tools for liver transplant decisions using data collected and tested in real-life conditions which opens the door for AI and machine learning to be used safely and effectively in healthcare decision-making.
If these new prediction tools prove to work well, they could be adopted by national Organ Sharing Organizations (OSOs) across Europe, which are all partners of the project, and support clinicians to make better decisions, improve survival chances for patients, and ensure fairer access to transplants regardless of country or hospital.
To make sure these tools can be used widely, we still need to:
• Run multi-country pilot studies to test model adoption in routine clinical practice and OSO decision-making workflows.
• Implement regulatory and standardization frameworks to facilitate the integration of AI tools into national transplant systems, ensuring compliance with GDPR.
• Access to financial support for the extension to non-EU countries, the sustainability and potentially commercialization of software tools and training modules developed within LEOPARD.